You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242516527

Reliability Analysis of a Helicopter Composite Armor using Stochastic


Optimization Techniques

Article

CITATIONS READS

0 67

2 authors, including:

A. B. Jorge
Universidade Federal de Itajubá (UNIFEI)
19 PUBLICATIONS   66 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Boundary element method applications to engineering problems and other numerical metthods. View project

All content following this page was uploaded by A. B. Jorge on 08 July 2014.

The user has requested enhancement of the downloaded file.


EngOpt 2008 - International Conference on Engineering Optimization
Rio de Janeiro, Brazil, 01 - 05 June 2008.

Reliability Analysis of a Helicopter Composite Armor using Stochastic Optimization


Techniques

V. C. Santos, A. B. Jorge

Institute of Mechanical Engineering, UNIFEI - Federal University of Itajubá


Av BPS, 1303, Itajubá, MG, Brazil - CEP 37500-903
valquiria.santos@gmail.com
ariosto.b.jorge@unifei.edu.br

1. Abstract
In this work, a reliability analysis of an aeronautical armor is proposed, using stochastic optimization techniques. The reliability of
the armor can be assessed by evaluating several types of failures when the armor is hit by a projectile. The possible armor failures
include projectile penetration, excessive displacement, excessive deformation, non-absorption of the projectile energy by the
structure, etc. Also, the armor should be designed in order to avoid any conflict with aircraft design restrictions, such as the
maximum allowed weight and the aircraft centroid position. Stochastic optimization refers to the minimization or maximization of an
objective function in the presence of randomness. Uncertainties can occur in the projectile parameters, such as its mass, velocity,
angle of incidence, and also in the armor parameters, such as its material and geometric parameters. Also, there is some uncertainty
in the armor boundary conditions, depending on the way the armor is installed in the aircraft structure. The optimization procedure
includes several constraint equations, such as, kinetic energy and maximum permissible deformation, maximum weight and
permissible displacement for the aircraft centroid. Some of these constraint equations are obtained through regression of data
obtained from numerical simulation. ANSYS/LS-DYNA® is used for the numerical simulation. This simulation is carried out using a
specially designed, existing code for non-linear transient dynamic events, such as ballistic impacts. To simplify the aircraft armor
analysis, the protected area is divided into a mesh, consisting of a set of rectangular plates, and only the impact in one plate is
considered. In this work, the armor is made of two thin layers. The first layer is made of a ceramic material while the second layer is
made of a composite material. The impact is assumed to occur in the center of the plate. Numerical results for the armor reliability
are obtained and compared for the various modes of failure that were assumed.

2. Keywords: Reliability Analysis, Stochastic Optimization, Ballistic Impact, Composite Materials.

3. Introduction

This study focuses a optimization reliability for an armor for a helicopter. The armor plate is projected for the helicopter floor and is
made of composite and ceramic materials. The combination of ceramic and composite materials for armor is still an ongoing
research, with few available published results. A composite is made of two types of materials: matrix and reinforcement. Often,
composites use fibers as reinforcement, such as graphite, glass or polymer fibers. The type of matrix and the shape of the fibers
provide different materials. For example, fibers can be put in only one direction in the material or can be woven or even cut and
randomly distributed in the matrix. The composite will have stronger mechanical properties in the direction of fibers and weaker in
any other direction.
The best ballistic performance with respect to impact parameters and also to physical properties of the projectile and of the armor
materials is chosen from the comparison between two optimization models, in this work. The model proposed maximizes the
reliability of an armor plate made of two materials. The minimum thickness for non-penetration of a projectile is observed for each
material combination, using numerical data obtained from several values of the plate thickness. The numerical code used to simulate
the dynamics penetration of the projectile considers failure modes leading to a breach in the armor plate when excessive
deformations or stresses are attained in the fiber or in the resin at the impact point. The optimization code considers constraint
equations, including restrictions for the position of the center of gravity of the helicopter, for the maximum allowed thickness of the
armor plate, and for the energy absorbed by the armor plate as a function of the kinetic energy of the projectile and displacement of
the plate of composite material.

2 Analytical model
2.1 Material Properties

The composite armor plate is formed by a combination of two materials with different properties. These materials act in a
complementary way during the projectile penetration process. Superior protection results are obtained when the first layer, the one
that receives the initial impact, is made of a fragile material, such as Alumina. This material has the function of destroying the
projectile top and also to dissipate most part of the projectile energy. The next layer is made of a ductile material that has the purpose
to absorb the residual energy of the fragments from the projectile and also from the armor material itself, by changing the kinetic
energy into plastic deformation energy (Ref. 1).
In this situation, where the ceramic material is placed above the composite material, the pressure distribution along the projectile is
less uniform than in the case where the projectile is impacting directly on the composite material itself. The plate made of composite
material reduces the pressure of the shock wave and retards the propagation of the corresponding front wave (Ref. 2). In this work, a
first plate is ceramic material and a second composite material was used, with material properties from table 1.

Table 1 - Material Properties (adapted from reference (Ref. 3)).


Properties Projectile Ceramic Plate Composite Plate
ρ [kg/m3] 7890 3720 1460
E X [GPa] 202 303 76
EY [GPa] - - 5.5
EZ [GPa] - - 5.5
υ XY 0.3 0.21 7.56E-02
υYZ - - 7.56E-02
υ XZ - - 6.98E-01
Gab , Gbc , Gca [GPa] - - 2.3
σ X [MPa] 792 2100 34
ET [GPa] 21 - -
C 40 - -
P 5 - -
Ef 0.15 - -
X T [Pa] - - 1400E+06
YT [Pa] 12 E+06
YC [Pa] 53 E+06
S [Pa] - - 34E+06

where E X is the young’s modulus, ρ is the material density, υ XY is the Poisson ratio, Gab , Gbc , Gca is the shear modulus, X T is
the longitudinal tensile strength, YT is the transverse tensile strength, YC is the transverse compressive strength and S is the shear
strength
The failure modes for the composite material depend on several parameters, such as the loading applied on the plate, the sequence of
piling up of the layers of composite material, the plate geometry and the mechanical and chemical properties of the layers (fiber,
matrix and interface). Some of these parameters can be more important than others, and different parameter combinations can
generate a great variety of failure modes.
In the impact problem, many parameters are involved (mass, speed, the final form, and kinetic energy of the projectile, geometry of
the plate, boundary conditions and characteristics of the materials) and depending on the interaction of these parameters different
answers are obtained. The material for the projectile is assumed steel 4340. At the present work, the results for composite materials,
observing the kinetic energy of the projectile and acceptable displacement of the plate, is made for each modification of the number
the layers and orientation for the composite plate.

2.2. Computational Simulation

A 3D finite element (FE) analysis model is created using ANSYS/LS-DYNA® to simulate the transverse impact of a projectile into a
patch of ceramic/composite material. Impact analysis involves the hit of a 7.62 mm diameter projectile into an armor plate whose
thickness ceramic plate is varied and then an investigation of the resulting penetration is performed. For the second plate (composite)
is varied number the layers and orientation. The first’s constant varied was:
1. thickness ceramic plate;
2. number the layers for the composite plate and;
3. orientation for the composite plate.
In order to analyse the effect caused at the material and the fracture models on the penetration of the projectile into the armor plate,
for the given impact velocity, FE simulations are conducted using a plastic kinematic model for the projectile and for ceramic plate,
and a damage composite material model for the composite plate (Ref.4).
Plastic kinematic hardening material model is a strain rate dependent elastic–plastic model. In this model, strain rate is taken into
account using the Cowper–Symonds model which scales the yield stress by the strain rate dependent factor as shown in (1):

⎡ 1

⎢ ⎛ε ⎞ ⎥
. p

σ y = ⎢1 + ⎜ ⎟ ⎥σ 0 (1)
⎢ ⎜⎝ c ⎟⎠ ⎥
⎢⎣ ⎥⎦

where σ 0 is the initial yield stress, ε is the strain rate, c and p are the Cowper–Symonds strain rate parameters. To allow crack
growth and fracture during penetration, the plastic kinematic hardening material model is coupled with an element-kill algorithm
available in ANSY/LS-DYNA® that removes the damaged elements from the mesh when the damage variable reaches the
predetermined critical value of the 0.15 (Ref. 4).
Composite Damage Model was developed by Chang and Chang (Ref. 5) for failures in the composite material. The material
parameters, such as, longitudinal tensile strength, transverse tensile strength, shear strength, transverse compressive strength, and
nonlinear shear stress parameter, are used in the three failure criteria. These five material parameters are obtained from strength
measurement of the material. Also, elastic modulus ( E X , EY , EZ ), ρ material density e Poisson ratios ( υ XY , υYZ , υ XZ ) are used at
the model. For the failure assumed Fcompression > 1 , the final failure model is due to fiber breakage is:
2
⎛σ ⎞ _
F fiber = ⎜ 1 ⎟ + τ (2)
⎝ S1 ⎠

The model used in this work for the numerical simulations assumes a rectangle plate with fixed boundaries. To reduce the
computational costs only half of the plate was simulated. Also, the model considers that, for an effective armor, the plate absorbs the
entire kinetic energy of the projectile. In future works, new considerations could be included in the model, such as changes in
projectile mass and speed, different angles of projectile incidence, and different boundary conditions, in order to consider more
realistic parameters.
Following the ballistic protection (Ref. 6), the armor plate was idealized as a 25 cm × 25 cm square plate, surrounded by a plate
region not deformed by the impact. This plate size is a standard measure and each plate is put on supports until the arrangement
completes the aircraft cockpit. Initial thickness value in the ceramic plate is given as a mean of 12 and a standard deviation 2 for the
ceramic plate and composite number the layers as a mean of 24 and a standard deviation 12. For the orientation the composite plate is
assumed angle the 0/90 and 0/45.
The simulation model was performed with parameters complying with the norm for Ballistic Resistant Protective Materials (Ref. 6).
Thus, according to this norm, the projectile was assumed, for these simulations, to be a 762-caliber of 4.7 × 10−3 kg mass, with speed
of 838 m / s . Also, the initial projectile speed detailed in the norm is higher than the speed at the moment of the impact, as the
projectile slows down during its trajectory, thus the value stated in the norm relates to the worst-case scenario. Also, in this work the
projectile is assumed to reach the armor plate perpendicularly, is shown in Figure 1. Thus, the influence of varying the projectile
angle of incidence is not evaluated in this work.

Figure 1. Model simulation armor

2.3. Meta-model

Normal distribution is assumed for this work. This model is a model widely used for the distribution of a random variable. Random
variables with different means and variances can be shaped by the density of probability functions normally, with appropriate choices
of the centre and the width of the curve. The rating N ( μ ,σ 2 ) is a normal distribution with mean μ and variance σ 2 . In the case of
this work we have as random variables X 1 (young’s modulus) and X 2 (longitudinal tensile strength), X 3 (composite plate
thickness).
In the manufacture of composite may has imperfections, assumed this work vary two properties that will be used in computational
simulation. We assumed, as fix the thickness of the plate ceramic. The composite material is formed of fibers and resin, this work
was assumed the material as isotropic. Initial tests showed that the orientation of the layer composite random variable was not as
significant as found in Figure 2. The initial test for the composite layer with the angle of 0/45 with μ = 45° e σ 2 = 3 × σ 2 . This
figure 2 show comparison of the kinetic energy for small deviations in the standard deviation no significance.
Comparison layers orientation

1,800E+13
1,600E+13
1,400E+13

Kinetic energy (Ec)


1,200E+13
Orientation 0/45
1,000E+13
Orientation 0/42
8,000E+12
Orientation 0/48
6,000E+12
4,000E+12
2,000E+12
0,000E+00
1 24 47 70 93 116 139 162
Tempo (s)

Figure 2. Comparison layers orientation

If there are tree number of numbers, the following situations are possible; valor minimum and maximum for the X 1 and X 2 . The
total number of possible states will be 2n . If there tree factors, the number of possible states is 23 = 8 , with 1 central point. Each
sample point preclude the occurrence of the others, thus, they are mutually exclusive. The ceramic plate thickness was assumed
7mm. The young’s modulus ( X 1 ) have μ = 76 × e 9 (real proprieties) and μ + 3 × σ with σ = 5% μ , the longitudinal tensile strength
( X 2 )have μ = 1400 × e 6 (real proprieties) and μ + 3 × σ with σ = 10% μ , and composite plate thickness ( X 3 ) have μ = 2.3 and
σ = 0.5 , the thickness each layers is 0.125 mm .
Determine the relation between response variable and variables of interest is a hard work. Experiments are performed to measure the
effects of one or more variables at the final response of a certain problem. The decision of which factors influence in a particular
process or product is often faced. To perform an experiment, some factors are selected and a multivariate regression technique is used
(by Minitab®). In this work, tree multivariate regressions are performed for kinetic energy of the projectile, maximum permitted
displacement of the armor plate and reliability.
The variables X 1 , X 2 , X 3 were normalized with respect to their mean values, in order to obtain coefficients of same order of

magnitude in the response surface equations obtained by means of a regression technique using the available data (Ref.7).
The kinetic energy values, the maximum displacement and rupture stress values were obtained from ANSY/LS-DYNA®. The study
of the relationships between the variables is made through a statistical regression technique. In this work was considered a correlation
coefficient that is a measure of strength of linear relationship between two variables (Ref. 8). This coefficient shows that a given
variance in a variable is explained through another variable under study, indicating how the model fits the data.

3. Optimization Model and Numerical Results


3.1 Sequential quadratic programming (SQP)

The resulting problem of the nonlinear kinetic energy objective function to be minimized, subject to nonlinear constraint equations
can be set in the general framework of a nonlinear programming problem. In this work, the sequential quadratic programming (SQP)
technique was used to obtain a local minimizer, i.e., a local solution for the optimization problem. SQP is a method that consists in
solving the following general problem:

min f ( x )
x ∈ Rn

Subject to g ( x ) ≤ 0 (3)
h( x) = 0
where
f ( x ) : R n → R , g ( x ) : R n → R md e h ( x ) : R n → R mi .

n – Number of variables;
md – Number of inequality constraints;
mi – Number of equality constraints.
The Lagrangian function for this problem can be written as:

L ( x, λ , π ) = f ( x ) − λ T h ( x ) − π T g ( x ) (4)

where λ and π are the vectors of the Lagrange multipliers associated with the equality and inequality constraints, respectively.
Using the conditions of first-order optimality:

∇L ( x, λ , π ) = 0 (5)

The solution of this problem can be obtained applying Newton’s method to “Equation (9)”. Thus, at iteration k, the following
equation is obtained

⎛ Δx k ⎞
⎜ ⎟
∇ L ( x , λ , π ) ⎜ Δλ k ⎟ = −∇L ( x k , λ k , π k )
2 k k k
(6)
⎜ Δπ k ⎟
⎝ ⎠

where Δλ e Δπ are the Lagrange multipliers. This way, with the first-order optimality conditions for the optimization problem, the
problem can be stated as

Δx ⎡⎣∇ xx L ( x k , λ k , π k ) ⎤⎦ Δx + ΔxT ⎡⎣∇ x L ( x k , λ k , π k ) ⎤⎦


1 T 2
min
2
Subject to ⎡⎣∇g ( x k ) ⎤⎦ Δx + g ( x k ) ≤ 0
T (7)

⎡∇h ( x k ) ⎤ Δx + h ( x k ) = 0
T

⎣ ⎦

One can observe that the system represented in “Equation (11)” consists of minimizing a quadratic function (quadratic approach for
the Lagrangian) subject to linear constraints.
In this way, through the SQP method, for each iteration, a quadratic program is solved. This method finds the Newton search
directions ( Δx , Δλ and Δπ ). They are used in the update of the values of x , λ and π , repeating the process for the new point. In
this work, a code was developed in MATLAB® to minimize the weight of the armor in the aircraft.

3.2 Numerical Results

Several energies are involved in the impact process. The first energy to consider is the projectile kinetic energy, which starts from its
initial value decreasing to its final value, after penetrating the armor and exiting on the other side of the armor plate, or decreasing to
zero, if the projectile is stopped by the armor plate. Other types of energy are related to the energy absorbed by the projectile in its
plastic deformation (mostly due to the ceramic part of the armor plate), or the energies absorbed by the armor plate in its own
deformation, either in the form of elastic potential energy of deformation. The energy can also be related to the energies absorbed by
the armor plate in the form of plastic energy of deformation, including here the destruction of the ceramic layer, or the rupture of the
fibers, or the breach of the resin, in the composite layer. Also, some amount of the projectile initial energy could be transferred to the
armor in the form of kinetic energy, but at its final position the armor kinetic energy goes back to zero. The amount of kinetic energy
lost by the projectile must be absorbed by its deformation, or by the armor deformation, or by the deformation of the aircraft structure
to which the armor is attached. In this work, only the total loss of the projectile kinetic energy is considered, and the fractions of
energy that go to each of these processes are not investigated (Ref 9).
The model used in this work for the numerical simulations assumes a rectangle plate with fixed boundaries. The computational cost
is reduced using half of the plate at the simulation. Thus, the remaining part of the armor plate is not influenced by the impact and is
assumed to remain motionless. Also, at the model is considered that, for an effective armor, the plate absorbs the entire kinetic
energy of the projectile. In future works, new considerations could be included in the model, such as changes in projectile mass and
speed, different angles of projectile incidence, and different boundary conditions, in order to consider more realistic parameters.
To obtain the kinetic energy equation (objective function), data from the ANSYS/LS-DYNA® simulation is function of the X 1 , X 2 , X 3
(8).

f1 = Ec=-78,3+183 X 1 -0,02X 2 -23,6 X 3 - 91,3 X 12 + 11 X 3 2 + 0,09X 1 X 2 - 0,086 X 1 X 3 - 0,164X 2 X 3 (8)

The parameter used is coefficient of determination R 2 ( adj ) that evaluate the adjustment of a regression model and explains 100%
of the variability of the data.
The first constraint equation is weight of an armor plate of mass m´ , f1 = m, .g .Assuming the acceleration of gravity g as constant, a
function for the mass of the armor plate can be set, as indicated in (9):

f 2 = ( 0.007 × 3720 + X 31460 ) A (9)

0.007 - ceramic plate thickness;


3720 - ceramic material density ;
X 2 - composite plate thickness;
1460 - composite material density;
A - armor plate area (width times length in this case of a rectangular plate).

For the equation of weight was assumed the maximum allowed for the armor plate is 40kg.

For the optimization model, the armor plate was idealized as a (1 × 1.5) m rectangular plate in the aircraft cockpit. This armor plate
is formed of small square plates of ( 25 × 25) cm where only one of these small plates was simulated. In order to guarantee some
degree of effectiveness in the protection of the aircraft crew, the armor plate was assumed to be at the helicopter floor. This armor
plate was assumed to be installed in a HELIBRAS-EUROCOPTER helicopter model SQUIRREL B2 AS350, which is a helicopter
model currently in use by several military and police operators.
The aircraft center of gravity (CG) must remain within operational limits, usually established by means of an analysis of stability and
control of the aircraft. The aircraft mass M changes while the fuel is burnt and the ammunition is used. In terms of variation of the
position of the aircraft center of gravity due to the presence of the armor plate, the critical situation occurs when the aircraft is light,
with a minimum of fuel and ammunition. In this work, this situation is assumed to be when the aircraft mass is 2000 kg. With respect
to the total mass of the aircraft, another constraining equation could have been established, with the critical situation occurring when
the aircraft is heavy, with a maximum of fuel and ammunition. This constraining equation was not considered in this work, as the
armor masses involved in this problem were small with respect to the total aircraft mass.
For every aircraft type, an envelope establishing operational limits for the aircraft CG and mass is presented in the aircraft
documentation available for the user, usually in the pilot manual. When the armor is placed below the pilot’s seat, the aircraft CG
must remain within the limits specified in this envelope. From the mass-CG envelope in the SQUIRREL pilot manual (Ref. 10), the
longitudinal CG of the aircraft is seen to vary between the forward limit of 3.17 m and the rear limit of 3.46 m, for a 2000 kg aircraft
mass. In this pilot’s manual, the origin (or reference) for the CG location is a point located 3.40 m in front of the center line of the
main rotor head. In this work, the origin of coordinates is assumed to be exactly below the head of the main rotor, corresponding to
the position of the aircraft CG without the armor plate. From this origin, two distances can be measured: X , the aircraft CG position
without the armor plate (hence zero), and X CG ′ , the position of the armor plate center being inserted in the aircraft (assumed to be
constant). The equation for X CG , the aircraft CG with the armor plate, leads to two constraint equations, one for each operational
limit of the longitudinal CG, the forward and the rear limits, now complying with the following operational interval:
( −0.23 ≤ X CG ≤ 0.06) m . The position of the aircraft CG is written in (10) as

X CG = ( X .M + X CG ′ .m′ ) ( M + m′ ) = ( X CG ′ .m′ ) ( M + m′ ) (10)

where:
X CG - position of the center of the armor plate (assumed constant);
M - aircraft mass without the armor plate (assumed in its critical case);
m´ - armor plate mass (varying with thickness and material density).

Substituting f1 from (3) into (4), X CG is obtained as in (11):

X CG = ( X CG ′ . f 2 ) ( M + f 2 ) (11)

The operational limits for the aircraft CG, as discussed above, lead to two constraining equations as in (6) and (7):

(X CG ′ . f 2 ) ( M + f 2 ) ≤ 0.06 (12)

(X CG ′ . f 2 ) ( M + f 2 ) ≥ −0.23 (13)

The fourth constraint equation is displacement; this was acquired of the simulation with ANSYS/LS-DYNA® as the displacement of
the element opposite to the projectile in the second plate (composite material). For this work, the maximum displacement is limited to
10 mm and this displacement is reached after 60 μ s .
Displ.=0,0335-0,068X 1+0,0141X 2 +0,0137X 3 + 0,0385 X 12 -0,0099 X 3 2 -0,0166X 1 X 2 +0,00382 X 1 X 3 (14)

The coefficient of determination R 2 ( adj ) that evaluate the adjustment of a regression model and explains 54.5% of the variability
of the data.

The last constraint equation is the rupture Stress (15):

σ rup =24,5-41,6X 1 -5,32X 2 -1,73 X 3 + 18,9 X 12 +0,709 X 3 2 +5,23X 1 X 2 -1,16X 1 X 3 +1,18X 2 X 3 (15)

The criterion assumed for the element opposite to the projectile in the second plate (composite material) is less that σ rup <1.10e9 Pa .
After rupture stress achieve this value to armor plate not considered effective. Finding the CG, weight, displacement and rupture
stress equation, and having already set the kinetic energy function to be minimized and the constraint equations, the model were
idealized for the optimization problem solved through MATLAB®. This problem was solved using Sequential Quadratic Programming
(SQP) that finds the local optimal. The model is represented below by (16):

min f1 = -78,3 + 183 X 1 - 0 ,02 X 2 - 23,6 X 3 - 91,3 X 12 + 11 X 32 + 0 ,09 X 1 X 2 - 0,086 X 1 X 3 - 0 ,164 X 2 X 3


x T Î T Rn

Subject to : Peso = f 2 = ( 0.007 × 3720 + X 31460 ) A ≤ 40kg


f1
X CG' - 0.06 ≤ 0
M + f1
f1
- X CG' - 0.23 ≤ 0 (16)
M + f1
Desl. = 0,0335 - 0,068 X 1 + 0 ,0141X 2 + 0 ,0137 X 3 + 0 ,0385 X 12 - 0 ,0099 X 3 2 - 0 ,0166 X 1 X 2 + 0,00382 X 1 X 3 ≤ 0.010 m
σ rup = 24 ,5 - 41,6 X 1 - 5,32 X 2 - 1,73 X 3 + 18,9 X 12 + 0 ,709 X 32 + 5,23 X 1 X 2 - 1,16 X 1 X 3 + 1,18 X 2 X 3 ≤ 1.10GPa

In this case, the optimal result for the kinetic energy function of the armor plate was obtained, for the different combinations of the
X 1 , X 2 and X 3 , as

X 1 = 7.6 GPa
X 2 = 1400 MPa (17)
X 3 = 1.3mm

Observing the results for SQP obtained from Matlab®, the optimal solution for the problem occurred directly depends on the initial
point. This fact is due to the SQP search a local optimum, this method indicated is global optimum as genetic algorithm. After several
run of SQP, getting several great, chosen to be the best of them as a better approach to a global minimum.
After the optimization using SQP was then evaluating the reliability of this result using a simplified technique response surface, was
assumed z1 be normal (to facilitate probability of z > 0 (success, or reliability) or 1 − Pfalha , or the probability of z < 0 , that is the
probability of failure of armor plate. Considering two variables (one relating to the demand on the system, load on the armor plate, S,
and the other to capacity of the system, resistance of the armor plate, R). Both S and R are random in nature; their randomess is
μ μ μ μ f (s)
characterized by their means S and R , standard deviations S and R , and corresponding probability density functions S and
f R (r )
(Ref 11,12).
The uncertainties in the load and resistance variables are expressed in the form of the probability density functions, we can express
the measure of risk in terms of the probability of the failure event or P( R < S ) as:
p f = P( failure ) = P ( R < S ) (18)
If both R and S are normal variables, that is, N = ( μR ,σ R ) and N = ( μ S ,σ S ) , then another random variable Z can be introduced as:
Z =R−S (19)

For an armor plate to provide protection, the projectile must not penetrate, and the initial kinetic energy of the projectile must be
completely absorbed by the armor, or by a combination of armor plate and aircraft structure, in the impact process. The numerical
simulations are computationally very extensive, and the time interval for the projectile kinetic energy to become zero can be very
large, leading to great computing times for the simulation of a complete stop of the projectile. In this work, a stop criterion was
adopted in the simulations to account for the amount of kinetic energy lost by the projectile during the impact. Thus, a criterion was
set so that if the projectile had lost or more of its initial kinetic energy after a time interval of after initial contact, the armor was
considered effective.
From the numerical simulations using ANSYS/LS-DYNA®, a set of data for the projectile kinetic energy while it penetrates the armor
plate was obtained, as a function of time, X 1 (young’s modulus) and X 2 (longitudinal tensile strength), X 3 (composite plate
thickness). The projectile kinetic energy at the time of 50 s was then noted. A multiple regression of the energy data, as a function of
X 1 , X 2 and X 3 was then performed, and a response surface was obtained as an equation. The assumed criterion is that the obtained
equation must be less or equal to 2% of the initial kinetic energy of the projectile to ensure non-penetration of the armor plate. The
regression analysis from a data set is a statistical technique to model and investigate a relation between variables.
z1 = 2% − f1
z1 = 2- (-78.3+183 X 1 -0.02X 2 -23.6X 3 - 91.3X 12 + 11X 3 2 + 0.09X 1 X 2 - 0.086X 1 X 3 - 0.164X 2 X 3 ) (20)
= 80.3 - 183 X 1+ 0.02X 2 +23.6X 3 + 91.3X 12 - 11X 3 2 - 0.09X 1 X 2 + 0.086X 1 X 3 + 0.164X 2 X 3

With the valor, X 1 , X 2 and X 3 same the μ1 , μ 2 and μ3 . The resulted is represented below by (20):

z1 = 1.38 (21)

The derivatives z1 with respect to, X 1 , X 2 and X 3 are, respectively:

⎛ ∂z1 ⎞
⎜ ⎟ = -183+182.6 X 1 - 0.09X 2 + 0.086 X 3
⎝ ∂x1 ⎠

⎛ ∂z1 ⎞
⎜ ⎟ = 0.02 - 0.09X 1+ 0.164X 3 (22)
⎜ ∂x 2 ⎟
⎝ ⎠

⎛ ∂z1 ⎞
⎜ ⎟ = 23.6 - 22 X 3 + 0.086 X 1+ 0.164 X 2
⎝ ∂x3 ⎠

μ z1 = μ ( 2 − f1 ) = 0,02 − μ f 1 = 1.38 (23)

2 2 2
⎛ ∂z1 ⎞ 2 ⎛ ∂z1 ⎞ 2 ⎛ ∂z1 ⎞ 2
σ z21 = σ 2f 1 = ⎜ ⎟ σ x1 + ⎜ ⎟ σ x2 + ⎜ ⎟ σ x 3 = 1.46 (24)
⎝ ∂x1 ⎠ ⎝ ∂x2 ⎠ ⎝ ∂x3 ⎠

The reliability of the results found with the method SQP was verified. This technique is a simplification of the response surface. The
response surface z1 was assumed normal to facilitate probability of z > 0 ⇒ reliability of armor plate.

5. Conclusions

This work presented an optimization procedure of an armor plate for a helicopter. As already mentioned previously, SQP method
looks for the local optimal, searched optimal composite plate thickness X 3 = 1.3 .The equations used in the optimization problem
were generated through a regression technique with data received from simulation with ANSYS/LS-DYNA®. The optimization
methods minimized the kinetic energy of the armor plate. Other restriction equations can be considered in the optimization problem
of the armor of a helicopter, such as, permissible maximum tension of the material and aircraft stability. The inclusion of these
equations in the optimization problem can improve the quality of the results. After de SQP optimization was done the reliability
theses results of the plate armor.

6. References

1. FAWAZ, Z.; ZHENG, W.; BEHDINAN, K., Numerical simulation of normal and oblique ballistic impact on ceramic
composite armours. Composite Structure, Vol. 63, pp. 387–395, 2004.
2. WILKINS, M. L., Mechanics of Penetration and Perforation. International Journal Engineering Science, Vol. 16, pp. 793-807,
1978.
3. BORVIK, T., HOPPERSTADA, O. S., BERSTADB, T., and LANGSETHA, M., Perforation of 12mm thick steel plates by
20mm diameter projectiles with flat, hemispherical and conical noses Part II: numerical simulations. International Journal of Impact
Engineering, Vol. 27, pp. 37–64, 2002.
4. ANSYS User’s Manual. Theory manual. ANSYS revision 5.2., 1995.
5. CHANG F. K.; CHANG K. Y. (1987). A Progressive Damage Model for Laminated Composites Containing Stress
Concentration. Journal of Composite Materials, v. 21, pp. 834-855.
6. National Institute of Justice, Ballistic Resistant Protective Materials – NIJ Standard 0108.01., 1981
7. S. S. RAO (1996): Engineering Optimization - Theory and Practice: 3rd Ed, Wiley.
8. BLISCHKE W. R., MURTHY D. N. P. Reliability – Modeling, Prediction, and Optimization. John Wiley & Sons, Inc, 2000.
9. KURTARAN H., BUYUK M., ESKANDARIAN A., Ballistic impact simulation of GT model vehicle door using finite
element method. Theoretical and Applied Fracture Mechanics, Vol. 40, pp. 113–121, 2003.
10. EUROCOPTER AS350 B2 Pilot Manual, Eurocopter, Marignane, France.
11. HALDAR A.; MAHADEVAN S. (2000). Probability, reliability and Statistical Methods in Engineering Design. John Wiley &
Sons, 1º ed, New York, 304 p.
12. S. S. Rao (1992): Reliability-Based Design, McGraw-Hill.

V i e w p u b l i c a t i o n s t a t s

You might also like